Detection of LEO Objects Using CMOS Sensor

Transcription

1 Trans. JSASS Aerospace Tech. Japan Vol. 14, No. ists30, pp. Pr_51-Pr_55, 2016 Detection of LEO Objects Using CMOS Sensor By Toshifumi YANAGISAWA, 1) Hirohisa KUROSAKI 1) and Hiroshi ODA 2) 1) Chofu Aerospace Center, JAXA, Chofu, Japan 2) Tsukuba Space Center, JAXA, Tsukuba, Japan (Received July 30th, 2015) We succeeded in detecting 10 cm LEO objects at 1000 km altitude with a CMOS sensor installed to the 18cm telescope by using fast frame rate of the CMOS sensor and FPGA-based image-processing technique. The LEO survey system using numerous sets of the CMOS sensor will contribute the monitoring LEO environments which is currently done with the space surveillance network of the United States and solving the space debris problem in the future. Key Words: Space Debris, LEO, CMOS Sensor 1. Introduction The space environment has recently deteriorated due to space debris, particularly in the low-earth orbit (LEO) region. 1) In order to protect active satellites in LEO, space debris must be precisely monitored. In this paper, monitor means detection of objects, determination of their orbits and maintaining them. Although currently, this is mainly handled by the Space Surveillance Network (SSN), via global radar observation sites of the United States, the SSN detection limit is about 10 cm which is insufficient to protect active satellites. Most satellites cannot survive a collision with an object of 1 cm. 2) Conversely, optical sensors are mainly used to monitor geostationary (GEO) orbits. 3) However, by enhancing the detection efficiency of optical sensors and PC performance, the extremely low cost of optical sensor compared to radar systems may enable optical sensors to monitor LEO objects like radars. The complementary metal oxide semiconductor (CMOS) sensors are rapidly replacing the charge-coupled device (CCD) sensors in the consumer camera market. With their significant technological improvements, these devices begin to compete with CCDs for scientific applications especially in the astronomical field. Schildknecht and his team discussed the application of CMOS for the observation of space objects like space debris. 4,5) We are developing CMOS sensors shown in Fig.1 which are designed for LEO objects observation. It can take 32 1K1K-frames of 10msec exposure with 2 seconds interval continuously. This is much faster than CCD sensor and a big advantage for the observation of fast moving LEO objects. We carried out LEO survey observations using the CMOS sensor. The data was analyzed using the field-programmable gate array (FPGA) based analysis system which was originally developed for the detection of small objects at GEO orbits and modified for the detection of LEO objects. As a result, we succeeded in detection of LEO objects with the size of about 10cm. This result means that CMOS sensor and FPGA based analysis system will be a powerful tool to monitor LEO Fig. 1. environments in the future. The detail of the CMOS sensor used in this study is shown in the section 2. The FPGA based analysis system is explained in the section 3. The survey and its results are described in section 4. The LEO survey system using a lot of CMOS sensors is proposed in the section CMOS Sensor CMOS sensor developed for LEO objects observation. A CMOS sensor designed for observations of LEO objects was developed by Nobuo Electronics co., LTD. 6) The main CMOS camera was ORCA-Flash4.0v2 manufactured by Hamamatsu Photonics co.,ltd. 7) It has 2K by 2K pixels with the 6.5 microns by 6.5 microns pixel size. 2 by 2 binning is possible for data saving. The quantum efficiency is about 30%, 65%, 73%, 65%, 43% and 20% at 400 nm, 500 nm, 600 nm, 700 nm, 800 nm and 900 nm, respectively. It can take 30 and 100 frames per second with the readout noise of 0.8 and 1.0 electron, respectively. The global positioning system (GPS) unit was installed to stamp the image acquisition time with msec order. Interval image acquisition is possible to prevent overflow of the data during long observation period. With a 18cm-telescope epsilon 180ED manufactured by Takahashi co., LTD 8) for the optics, the field of view of the sensor is Copyright 2016 by the Japan Society for Aeronautical and Space Sciences and ISTS. All rights reserved. Pr_51

2 Trans. JSASS Aerospace Tech. Japan Vol. 14, No. ists30 (2016) Fig. 2. The FPGA board manufactured by Soliton Systems co., LTD. 1.5-degree by 1.5-degree. 3. FPGA Based Analysis System Basic algorithm of the FPGA based analysis system is the stacking method. It uses multiple CCD images to detect very faint objects that are undetectable in a single CCD image. It stacks those images with presumed target motion to improve the signal-to-noise ratio by calculating the median. Details of the stacking method are shown in other papers. 9-11) The only weakness of the stacking method is the time required to detect an unseen object whose movement is not known, because a range of likely paths must be assumed and checked. For example, the analysis time for 65,536 processing iterations of 32 1,024 1,024-pixel frames, which are intended to detect objects moving within a pixel area, is about 280 hours using a normal desktop computer, which is not really practical. Although cataloged space debris whose motion can be estimated in some way are easy to detect, finding un-cataloged space debris is realistically impossible. Although using many computers in parallel to reduce the analysis time may be a solution, considering upcoming large-format CCD cameras and CMOS sensors, a more sophisticated algorithm which greatly reduces the analysis time is needed to overcome this situation. We found that binarizing original images reduces analysis time dramatically and sum can be used instead of median, which is most time-consuming part of the stacking method, to get almost same results. Besides, this new algorithm is very simple, so that we can embed it to an FPGA, which is a programmable hardware device suitable for rapidly executing simple calculations. 12) Figure 2 shows the FPGA board manufactured by Soliton Systems co., LTD. 13) Implementing the algorithm in hardware reduces the analysis time by 1,200 times over the original stacking method implemented in software. In our previous work, we used the FPGA based analysis system for the detection of small GEO objects by analyzing CCD frames. 12) The typical readout time of CCD cameras is a few second, which is enough for the observation of GEO objects. However, observation of LEO objects requires much faster readout time. Recently, improvements of CMOS sensors are remarkable. The most promising advantage of a CMOS sensor over a CCD camera is the readout time. It can read 100 frames per second. This is the big advantage for the Fig. 3. The faintest object detected in this study. Top figure shows the original image around the detected object. The second and the third images are the stacked images using 4 and 8 frames, respectively. The bottom image is the final stacked image using 32 frames. observation of LEO objects. In order to detect un-cataloged LEO objects, one of the effective observation methods is to stare one sky region with optical sensors and wait for the LEO objects passing through the field of view (FOV) of the sensors (a few degrees). In this method, optical sensors take images as fast as they can. Since the typical passing time of LEO objects is about 10 seconds, CCD cameras can only take a few frames, whereas, CMOS sensors can take hundreds. We can apply the FPGA based analysis system to the frames taken with CMOS sensors to detect faint LEO objects. Another advantage of CMOS sensors is the readout noise. As described in Section 2, readout noise of the CMOS sensor of this study is 1.0 electron, whereas, typical readout noise of CCD camera is about 10.0 electrons. In the case of long time exposure where the sky background noise is dominant, 10.0 electrons of CCD cameras may be negligible. However, in the case of LEO objects observation where extremely short exposure time like a few msec is required, 1.0 electron of CMOS sensors will greatly contribute to the noise suppression. 4. Survey Observation of LEO Objects Survey observations using the CMOS sensor described in Section 2 were carried out on November 22, December 25, 26 of 2014 and February 20 of The CMOS sensor was installed to the 18cm-telescope described in Section 2. In each day of 2014, 1800 sets of 32 frames of 10m-second exposure with 2-second interval were taken for 1 hour sets were taken for 2.5 hours on February 20 of All the data was analyzed using the FPGA based analysis system described in Section 3 offline. Analysis time for one set is about 6.6 minutes. As 3 FPGA boards were used simultaneously, analysis time of 1800 sets data (for one hour) was 66 hours. Since the amount of data from the CMOS sensor is about 100 Pr_52

3 T. YANAGISAWA et al.: Detection of LEO Objects Using CMOS Sensor Table. 1. The detail of the detected LEO objects. date Time(UT) Name(Satellite number) Magnitude(V) 2014/11/22 18:42:46.3 SL-8 R/B (09598) /11/22 18:50:54.3 BREEZE-M DEB (37198) /11/22 19:15:32.3 NOAA 3 (06920) /11/22 19:16:50.3 CZ-4B R/B (25732) /12/25 19:40:58.3 Not Identified /12/25 19:44:34.3 WESTFORD NEEDLES (20005) /12/25 19:53:42.3 COSMOS 1041 (11050) /12/25 20:09:20.3 COSMOS 928 (10141) /12/26 09:09:29.3 COSMOS 1925 (189838) /12/26 09:14:34.3 Not Identified /12/26 09:14:52.3 COSMOS 1110 (11425) /12/26 09:21:30.3 COSMOS 1162 (11697) /12/26 09:51:50.3 GLOBALSTAR M073 (37913) /02/20 10:19:28.3 Not Identified /02/20 10:24:00.3 BREEZE-KM R/B (39060) /02/20 10:26:08.3 THORAD AGENED DEB (04155) /02/20 10:56:52.3 Not Identified /02/20 10:59:02.3 METEOR 3-3 (20305) /02/20 11:10:52.3 Not Identified /02/20 20:29:30.3 Not Identified /02/20 20:35:22.3 Not Identified /02/20 20:42:30.3 Not Identified 6.55 Fig. 4. The brightness distribution of detected objects with the CCD sensor. X- and y-axes show the brightness in magnitude and number, while the blue and red columns represent cataloged and un-cataloged objects, respectively. The presumed sizes calculated under conditions of 1000km altitude, diffuse reflection with a 90-degree phase angle and albedo 0.1 were shown for 8-, 10-, and 12-magnitude. times of CCD, the FPGA based analysis system originally developed for the analysis of GEO data is not good enough for the analysis of LEO data. We need to reduce analysis time to about 12 hours at least to track the target next day. As a result of analysis, 22 objects were detected. Table 1 shows the detail of the detected objects. The faintest object is about 13.5 magnitude which is about 10cm objects at 1000km under the assumptions of diffuse reflection and albedo of 0.1. Figure 3 shows the original images and the stacked image of the faintest objects. In the previous work, the CCD sensor was used for LEO Fig. 5. The brightness distribution of detected objects with the CMOS sensor in this study. X- and y-axes show the brightness in magnitude and number, while the blue and red columns represent cataloged and un-cataloged objects, respectively. survey observations with the same optics and 169 objects were detected in 16 days survey observations. 14) Figure 4 shows the brightness distribution of detected objects with the CCD sensor. Here, the X- and Y-axes show brightness in magnitude and number, and the blue and red columns represent cataloged and un-cataloged objects, respectively. The presumed sizes calculated under a condition of 1000km altitude, diffuse reflection with a 90-degree phase angle and albedo 0.1 were shown at 8-, 10-, and 12-magnitude in the figure. Since the CCD could not take many frames as compared with the CMOS sensor, FPGA based analysis system could not be used. As shown in Fig.4, the detection limit of the previous work done with the CCD sensor was about 30 cm. Even with the large detection limit, about 15% of detected objects were Pr_53

4 Trans. JSASS Aerospace Tech. Japan Vol. 14, No. ists30 (2016) Fig. 6. un-cataloged. The detection limit of the SSN is 10cm in the LEO. The discrepancy may be caused by the difference between the optical reflection properties and radar ones of the LEO objects. Figure 5 shows the brightness distribution of detected objects with the CMOS sensor in this study. Although the number of detected objects is small, which is caused by small number of observation date and narrower FOV of the CMOS sensor, fainter objects were detected which attributes to the usage of the FPGA based analysis system. Un-cataloged ratio increased to 36%. This phenomena was also confirmed by another study using the NASA 3.0 m diameter Liquid Mirror Telescope. 15) It showed that a quarter of LEO objects exceeding 30 cm and a half of those exceeding 10 cm are uncatalogued which means the amount of uncatalogued objects exceeding 10 cm is almost equivalent to the catalogued total. From these facts, using CMOS sensor for observations of LEO objects, which enable us to take numerous frames and use FPGA based analysis system, has a great possibility for the future LEO observation network. 5. LEO Optical Survey System Using a Lot of CMOS Sensors We are proposing LEO optical survey system using a lot of optical sensors as shown in Fig. 6 to complement current radar system for LEO monitoring and the space situation awareness. 16) The observation site is placed around the center of the left figure and each quadrangular prism pinned to the site represents the FOV of each optical sensor. Colorful lines circulating the Earth represent trajectories of LEO objects. Many LEO objects will traverse the two fan-shaped fences comprising optical sensors, which will help obtain the long arc of those objects and their precise orbital determination. For optical observation, the site must be in the shadow of the Earth and LEO objects not, which are conditions realized a few hours before sunrise and after sunset. The previous study showed observations of two consecutive passes performed at two longitudinally separate sites, whereby two sets of observation separated by 60 degrees at each site allowed us to determine the orbit precisely. 17) Since the typical orbital period LEO optical observation system using a lot of CMOS sensors. is about 100 minutes, the ideal longitude separation of both sites is 25 degrees, which is equivalent to the Earth motion of that period. By using 40 sets of a CCD sensor and a 18cm telescope at each site, about 60% of LEO objects larger than 30cm will be detected and orbit-determined in four months. In this calculation, we assumed that the weather was clear for four months. Two tracking sites which are placed to both polar regions are needed to maintain orbital elements of those objects. The detail of the system is described in Yanagisawa et al. 18) As discussed in the section 4, with these optical settings, about 15-25% of detected LEO objects will be un-cataloged. This will help to reduce the risk of collisions between those objects and active satellites. By replacing CCD sensors to CMOS sensors, detection limit will improve to 10cm. As also discussed in Section 4, the setting will increase un-cataloged ratio up to 36-50%. This means that the number of cataloged objects will increase by a factor of when the proposed system is used along with the SSN of the United States. 6. Conclusions By enhancing the detection efficiency of optical sensors and PC performance, the extremely low cost of optical sensor compared to radar systems may enable them to monitor LEO objects like radars. Basic LEO debris observation unit consisting of optical CMOS sensor and FPGA based analysis system was developed. LEO survey observations were carried out using this unit and the result showed that the unit can detect about 10cm LEO objects at 1000km altitude. We need to reduce analysis time for practical use but the result is promising. The LEO survey system using numerous sets of the unit will contribute the monitoring LEO environments, which is currently done with the Space Surveillance Network of the United States. References 1) Liou, J.-C. and Johnson, N.L.: Risks in Space from Orbiting Debris, Science, 311 (2006), pp ) Christiansen, E. and Kerr, J.: Ballistic Limit Equations for Spacecraft Shielding, Int. J. Impact Eng, 26 (2001), pp ) Schildknecht, T., Musci, R., Ploner, M., Beutler, G., Flury, W., Kuusela, J., de Leon Cruz, J. and de Fatima Dominguez Palmero, Pr_54

5 T. YANAGISAWA et al.: Detection of LEO Objects Using CMOS Sensor L.: Optical Observations of Space Debris in GEO and in Highly-eccentric Orbits, Adv. in Space Res., 34 (2004), pp ) Schildknecht, T., Hize, A., Schlatter, P., Silha, J., Peltonen, J., Santti, T. and Flohrer, T.: Improved Space Object Observation Techniques Using CMOS Detectors, Proceedings of 6 th European Conference on Space Debris, Darmstadt, Germany, ) Silha, J., Schildknecht, T. and Flohrer, T.: Improved Space Object Orbit Determination Using CMOS Detectors, Proceedings of 2014 AMOS Technical Conference, ) (accessed May 18, 2015). 7) (accessed May 18, 2015). 8) (accessed May 18, 2015). 9) Yanagisawa, T., Nakajima, A., Kimura, T., Isobe, T., Futami, H. and Suzuki, M.: Detection of Small GEO Debris by Use of the Stacking Method, Trans. Japan Soc. Aero. Space Sci., 44 (2002), pp ) Yanagisawa, T., Nakajima, A., Kadota, K., Kurosaki, H., Nakamura, T., Yoshida, F., Dermawan, B. and Sato, Y.: Automatic Detection Algorithm for Small Moving Objects, Publ. Astron. Soc. Japan, 57 (2005), pp ) Nakajima, A., Yanagisawa, T., Kimura, T., Isobe, T., Tsujii, T., Yamamoto, M., Hoshino, T., Suzuki, M. and Futami, H.: Space Debris Observation by Ground-Based Optical Telescope, Proc. 22nd Int. Symp. Space Technology and Science (2000), pp ) Yanagisawa, T. and Kurosaki, H.: Detection of Faint GEO Objects Using JAXA s Fast Analysis Methods, Trans. JSASS Aerospace Tech. Japan, 10, ists28 (2012), pp. Pr_29-Pr_35. 13) (accessed May 18, 2015). 14) Yanagisawa, T., Kurosaki, H. and Oda, H.: Ground-based Optical Observation System for LEO Objects, Adv. in Space Res., 56 (2015), pp ) Mulrooney, M.: The NASA Liquid Mirror Telescope, Orbital Debris Quart. News, 11, 2 (2007), pp ) Yanagisawa, T., Kurosaki, H. and Oda, H.: Investigation of Ground-based Optical Observation System for LEO Objects, Trans. JSASS Aerospace Tech. Japan, 12, ists29 (2014), pp. Pr_41-Pr_46. 17) Technical Material of JAXA, QNX , ) Yanagisawa, T., Kurosaki, H., Oda, H. and Tagawa, M.: Investigation of Ground-based Optical Observation System for LEO Objects, Proceedings of 58th Space Sciences and Technology Conference, JSASS , 2014 (in Japanese). Pr_55